Debug logging on linux Worker/Nuke

[drtic_cz]

Hi, I'm trying to get working Qube 5.5 worker on Linux with Nuke 6.1v1. Unfortunately, jobs are still failing without any message in stdout, stderr or in /var/log/workerlog.
Is there any possibility to run worker or Nuke with more detailed logging? There is nothing useful in i.e. $worker --help..

Thanks

[jburk]

If you look in the qubeGUI's "properties" pane when you select your nuke job, you will see the nuke command-line which is being run on the workers.  It will probably contain tokens that look like QB_FRAME_NUMBER.

Log into one of the workers, and cut&paste this command into a command-prompt (didn't say if you're on Windows / OS X/ linux), replacing QB_FRAME_NUMBER with a valid number in the frame range. 

There may be some error output that will shed some light on why these are failing so quickly.

It the command runs successfully, it may be due to the job's execution user lacking permission.

You may need to determine who the job is running as; if your workers have proxy_execution_mode set to "proxy", you need to log in as the user defined in "proxy_account".  If the proxy_account is the default value of "qubeproxy", you will find the password for that user in the Qube Administration Guide.  hint: search the Admin Guide for "$$"

[drtic_cz]

Ok, that sounds logic. I tried it before, but I missed the qb_frame arguments.. So I tried it now. Btw worker is on Linux, submitter on windows. Cmdline command from the qubegui is:

Code: [Select]

"Nuke6.1" -t -x  -- "//server/share/folder/project/script.nk" QB_FRAME_START,QB_FRAME_END,QB_FRAME_STEP
Output from the worker shell:

Code: [Select]

Can't read //server/share/folder/project/script.nk No such file or directory
There are no Write operators in this script

But I thought, that the worker will try to resolve UNC path's somehow like this:

/server/share/folder..etc

So what I'm doing wrong? 

[jburk]

If you always submit from one OS to another, there is a "path translation map" available in the QubeGUI->Prefs.  Any path translations you define here get applied to all jobs you submit.

These preferences are on a per-user basis.  If you need something a bit larger-scale, you can use symlinks on the linux hosts to help the linux host make sense of the UNC paths:

If the UNC path is: //server/share/folder

and you nfs-mount server:share on /mnt/share, you can make a symlink in the root of the linux drive:

mkdir /server; ln -s /mnt/share /server/share

so that if you log into the linux host and run

ls //server/share/folder

it will follow the symlink, traverse the nfs-mount, and file the files.

linux/unix/OSX don't care about the leading double-slashes.

[drtic_cz]

Ok, I have checked and repaired symlinks on the worker. Now, I have on worker /server/share/folders... So it is same as UNC from qubegui submitter and still failing without output.
If I try to run command from the cmdline, here is the output:

Code: [Select]

$ "Nuke6.1" -t -x -X Write1 -- "/server/share/folder/render_Qube.nk"  1,75,1
Nuke 6.1v1, 64 bit, built Jul 28 2010.
Copyright (c) 2010 The Foundry Visionmongers Ltd.  All Rights Reserved.
[20:02.49] ERROR: Read1: /PROJECTS/.../.../.../something.dpx: No such file or directory
Total render time: 0.00 seconds
Read1: /PROJECTS/.../.../.../something.dpx: No such file or directory

This kicked me to analyze the render *.nk script file and yes, there are absolute path's from mounted windows drive. So I suppose, that there is only one way to solve - Path translations per user, am I right?

[jburk]

You can define a global QubeGUI prefs file; refer to section 4.9.1 "Studio Preferences" in the Qube User's Guide.

4.9.1 Studio Preferences
One can set up default preferences used by the entire studio. The default preferences will be loaded first, and then the user preferences will override those values if non-default.

To save off a Preferences file based on the current preferences of the QubeGUI, select